Multi-time-horizon Solar Forecasting Using
Recurrent Neural Network
Sakshi Mishra
Transmission Planning Engineer, West Transmission Planning
American Electric Power
Tulsa, USA
sakshi.m@outlook.com
Abstract— The non-stationarity characteristic of the solar
power renders traditional point forecasting methods to be less
useful due to large prediction errors. This results in increased
uncertainties in the grid operation, thereby negatively affecting
the reliability and increased cost of operation. This research
paper proposes a unified architecture for multi-time-horizon
predictions for short and long-term solar forecasting using
Recurrent Neural Networks (RNN). The paper describes an end-
to-end pipeline to implement the architecture along with methods
to test and validate the performance of the prediction model. The
results demonstrate that the proposed method based on the
unified architecture is effective for multi-horizon solar
forecasting and achieves a lower root-mean-squared prediction
error compared to the previous best performing methods which
use one model for each time-horizon. The proposed method
enables multi-horizon forecasts with real-time inputs, which have
a high potential for practical applications in the evolving smart
grid.
Keywords— Artificial Neural Network, Forecasting, Predictive
Analysis, Recurrent Neural Network, Renewable Energy, Solar
Power, Multi-time-horizon Solar Forecasting, Smart Grid
I. INTRODUCTION
Today’s power grid has become dynamic in nature mainly
because of three changes in the modern grid: 1. Higher
penetration level of renewables, 2. Introduction and rapidly
increasing deployment of storage devices, and 3. Loads
becoming active by participating in demand response. This
dynamic modern grid faces the challenge of strong fluctuations
due to uncertainty. There is a critical need of gaining real time
observability, control, and improving renewable generation
forecast accuracy to enhance the resiliency and to keep the
operational costs sustainable. Independent system operators
(ISOs) have already been facing challenges with higher
renewable penetration on the grid due to the uncertainties
resulting from short-term forecasting errors. In the year 2016,
California ISO doubled its frequency regulation service
requirements, causing a sharp rise in the cost of requirements,
to manage the recurring short-term forecasting errors in
renewable generation [1]. The Western Electricity
Coordinating Council (WECC) could save $5 billion per year
by integrating wind and solar forecasts into unit commitment,
according to the study conducted by Lew et al [2]. Thus, it is
clear that the increased grid penetration levels of solar with its
inherent variability caused by a combination of intermittence,
high-frequency and non-stationarity, poses problems for grid
reliability and increases the grid operation costs at various
978-1-4799-7312-5/18/$31.00 ©2018 IEEE
Praveen Palanisamy
Alumni, Robotics Institute
Carnegie Mellon University
Pittsburgh, USA
praveen.palanisamy@outlook.com
time-scales. For example, day-ahead solar forecast accuracy
plays a significant role in the effectiveness of Unit
Commitment (UC); very-short-term solar forecast errors due
to fluctuations caused by the passing clouds lead to sudden
changes in PV plant outputs that can cause strain to the grid by
inducing voltage-flickers and real-time balancing issues. Thus,
solar power generation forecasting is an area of paramount
research, as the need for robust forecast for all timescales
(weekly, day-ahead, hourly and intra-hour) is critical for
effectively incorporating the increasing amount of solar energy
resources at a global level and contributing to the evolution of
the smart grid. Moreover, improving the accuracy of solar
forecasting is one of the low cost methods for efficiently
integrating solar energy into the grid.
The rest of the paper is organized as follows. The literature
is reviewed and the significant shortcomings of the current
forecasting approaches are recognized in Section II. Section II
further introduces the capabilities of the proposed unified
architecture and the novel algorithm to fill in the gap between
the need to improve the forecasting techniques and the existing
approaches. Section III introduces the proposed unified
architecture based on RNNs and the training procedure used
for implementing the neural network. Exploratory data
analysis, evaluation metric and the structure of the input data,
and the proposed algorithm are presented in Section IV.
Section V discusses the results and their interpretation. The
paper is concluded with Section VI, which also identifies the
future avenue of research in this method of solar forecasting.
II. LITERATURE REVIEW AND PROBLEM IDENTIFICATION
Forecasting methods which have been used for renewable
generation and electric load forecasting prediction can be
mainly classified into five categories: 1) Regressive methods,
such as Autoregressive (AR), AR integrated moving average
(ARIMA), and exponential smoothing (ES) models [3], [4],
[5], nonlinear stationary models; 2) Artificial Intelligence (AI)
techniques, such as Artificial Neural Networks (ANN) [6]-
[10], k-nearest neighbors [11]-[14], fuzzy logic systems
(FLSs) [15]-[17]; 3) Numerical Weather Prediction (NWP)
[18]; 4) Sensing (remote and local) [19]. 5) Hybrid models,
such as neuro-fuzzy systems [20]-[21], ANN and satellite
derived cloud indices [22], to name a few.
Numerical Weather Prediction (NWP) models are based
on physical laws of motion and thermodynamics that govern
the weather. For the places where ground data is not available,
18
NWP models are powerful tools to forecast solar radiation.
However, they pose significant limitations in predicting the
precise position and extent of cloud fields due to their
relatively coarse spatial resolution. Their inability to resolve
the micro-scale physics associated with cloud formation
renders them with relatively large error in terms of cloud
prediction accuracy. In order to mitigate this limitation, NWPs
are simulated at regional level (called Regional NWP) models,
downscaled to derive improved site-specific forecasts. NWP
has another limitation of temporal-resolution. The timescale of
output variables of NWP models is from 3 hour - 6 hours for
the Global Forecast System (GFS) and 1-hour (for mesoscale
model), which is not useful for predicting the ramp-rate and
very-short-term output fluctuations.
For the areas where the previous ground-based
measurement are not available, satellite based irradiance
measurement proved to be a useful tool [22]. The images from
satellite are used to analyze the time evolution of air mass by
the superimposition of images of the same area. Radiometer
installed in the satellite records the radiance, states of the
atmosphere (clear sky to overcast) impacts the radiance.
Satellite sensing has the main limitation of determining an
accurate set point for the radiance value under clear sky
conditions and under dense cloudiness condition from every
pixel in every image. Another limitation of solar irradiance
forecasting using remotes sensing with satellite is the
algorithms that are classified as empirical or statistical [23]-
[24]. These algorithms are based on simple statistical
regression between surface measurements and satellite
information and do not need accurate information of the
parameters that model the solar radiation attenuation through
the atmosphere. So the ground-based solar data is required for
these satellite statistical algorithms anyway.
The aforementioned limitations of NWP and sensing
models have steered the short-term solar forecasting research
towards time-series analysis using statistical models and more
recently AI techniques. Statistical techniques can mainly be
classified as [25]: 1) Linear stationary models (Autoregressive
models, Moving Average models, Mixed Autoregressive
Moving Average Models, and Mixed Autoregressive Moving
Average models with exogenous variables); 2) Nonlinear
stationary models; 3) Linear non-stationary models
(Autoregressive integrated moving average models and
Autoregressive integrated moving average models with
exogenous variables). Though these conventional statistical
techniques provide a number of advantages over NWP and
sensing methods, but these are often limited by strict
assumptions of normality, linearity, variable independence.
Artificial Neural Networks (ANN) are able to represent
complex non-linear behaviors in higher dimensional settings.
When exogenous variables like humidity, temperature and
pressure are considered in the process of solar forecasting -
ANNs act as universal function approximators to model the
complex non-linear relationships between these variables and
their relationship with the Global Horizontal Irradiance (GHI).
An ANN with multiple hidden layers can be called A Deep
Neural Network (DNN). With the advancements in
19
computational capabilities, DNNs have proven to be effective
and efficient in solving complex problems in many fields
including image recognition, automatic speech recognition
and natural language processing etc [26]. Although, feed-
forward neural network models have been used for solar
forecasting problem, the use of Recurrent Neural Networks
(RNN) models have not been explored yet, to the best of the
author’s knowledge. RNN is a class of ANN that capture the
dynamics of sequences using directed cyclic feedback
connections [27]. Feedforward neural networks rely on the
assumption of independence among the data points or
samples. The entire state of the network is lost after
processing each data point (sample). Unlike vanilla
feedforward neural networks, recurrent neural networks
(RNNs) exhibit dynamic temporal behavior by using their
internal memory to process arbitrary sequences if inputs,
which can be harnessed in predicting the irradiance for the
next time step by considering the input from many previous
times steps. Recent advances in parallelism, network
architectures, optimization techniques, and graphics
processing units (GPUs) have enabled successful large-scale
learning with RNNs overcoming their traditional limitations of
being difficult to train due to having millions of parameters.
Several methods have been proposed for solar forecasting
in the past but most of them were modeled for a particular
time-horizon and no single model performed well compared to
others for multi-time-horizon forecasting/prediction. In
addition, the state-of-the-art methods used for solar
forecasting primarily focuses on averaged rather than
instantaneous forecasts. This paper proposes two approaches
using RNNs. 1). A single system that is capable of being
trained to output solar forecast for 1-hour or 2-hour or 3-hour,
or for 4-hour time horizons. ii) A unified architecture that
can predict/forecast the solar irradiance for multi-time-
horizons; for example, the trained model can predict/forecast
the solar irradiance values for the 1-hour, 2-hour, 3-hour and
4-hour time horizons. Our proposed method is capable of
taking a time-series data as the input and provides predictions
with a forward inference time in the order of milliseconds,
enabling real-time forecasts based on live measured data. This
offers a great value for industrial applications that require real-
time multi-time-horizon forecasting for overcoming the
current operational challenges with high penetration of
renewable source of energy.
III. ARCHITECTURE AND ALGORITHM
The RNN resembles a feedforward neural network except
for additional directed edges. These edges span adjacent time
steps, introducing the notion of temporal component to the
model. Theoretically, the RNN architecture enables the
network to make use of past information in sequential data.
A. Recurrent Neural Network – Unified Architecture
The input to an RNN is a sequence, and its target can be a
sequence or a single value. An input sequence is denoted by
(x(1), x(2),...x(T)), where each sample/data-point x(t) is a real
valued vector. The target sequence is denoted by (y(1),
y(2),...y(T)) and the predicted target data-point is denoted by
(y(1), y(2),...y(T)). There are three dimensions to the input of the
RNN (shown in Figure 1): 1) Mini-batch Size; 2) Number of
columns in the vector per time-step; and 3) Number of time-
steps. Mini-batch size is the sample length (data-points in the
time-series). Number of columns are the input features in the
input vector. The number of time-steps is the differentiating
factor of RNN, which unfolds the input vector over time.
In a typical multilayer feedforward neural network, the
input vector is fed to the neurons at the input layer, which then
gets multiplied by the activation function to produce the
intermediate output of the neuron, this output then becomes
the input to the neuron in the next layer. The net input
(denoted by input_sumi) to this neuron belonging to the next
layer is the weight on connections (W) multiplied by previous
neuron’s output with the bias term, as shown in Equation 1.
An activation function (denoted by g) is then applied to the
input_sumi to produce the output from the neuron Equation 2
and 3.
݅݊‫݉ݑݏ̴ݐݑ݌‬௜  ൌ  ܹ௜  ή  ࢞࢏ ൅ ܾ (1)
ܽ௜  ൌ ݃ሺ݅݊‫݉ݑݏ̴ݐݑ݌‬௜ ሻ (2)
ܽ௜  ൌ ݃ሺܹ௜  ή  ࢞࢏ ൅ ܾሻ (3)
Figure 1 Input representation for the Recurrent Neural Network
For RNN network, at time t, neurons with recurrent edges
receive input from the current sample x(T) and also from
hidden node values h(t-1) in the network’s previous state
(Equation 4). Given the hidden node values h(t) at time t, the
output y(T) at each time t is calculated (Equation 5).
ࢎሺ‫ݐ‬ሻ  ൌ ߪሺࢃ௛௫ ‫ ݔ‬ሺ௧ሻ  ൅  ࢃ௛௛ ࢎሺ௧ିଵሻ  ൅  ܾ௛ ሻ (4)
ෝሺ࢚ሻ  ൌ ݃ሺࢃ ௬௛ ݄ሺ௧ሻ  ൅  ܾ௬ ሻ
࢟ (5)
where, Whx is the conventional weight matrix between the
input and the hidden layer and Whh is the recurrent weights
matrix between the hidden layer and itself at adjacent time
20
steps. bh and by are bias parameters. The proposed architecture
uses Rectified Linear Units (ReLU) as the activation function.
The network unfolds the given input at time t as shown in
Figure 1.
B. Algorithm description
The unfolded network is trained across the time steps
using an algorithm called backpropagation through time
(BPTT) [28]. The loss function used for this regression
problem is Mean Squared Loss (MSE). The loss function finds
the error between the target output and the predicted output
from the network. Gradients are computed using back-
propagation-through time [28] and the stochastic gradient
descent optimizer is used to update the weights so as to
minimize the loss. The RMSE is calculated for benchmarking
purposes.
The motivation to use a RNN is to identify and learn the
complex relationship between sequences of various exogenous
variables and their combined impact on the solar irradiance.
This, in the author’s view, enables the algorithm to recognize
non-linear contributing factors for example the atmospheric
conditions, which may lead to cloud formation in nearby time
horizon. This is one of the reasons the prediction RMSE is
lower in the proposed approach compared to other reported
approaches.
C. Solar Forecasting – input features and
predictions
The algorithm and the unified architecture developed in this
paper were trained and tested on data from the NOAA‘s
SURFRAD [31] sites similar to the previous works in the
literature [29][32]. The input features are: downwelling global
solar (Watts/m^2), upwelling global solar (Watts/ m^2), direct-
normal solar (Watts/ m^2), downwelling diffuse solar (Watts/
m^2), downwelling thermal infrared (Watts/ m^2),
downwelling IR case temp. (K), downwelling IR dome temp.
(K), upwelling thermal infrared (Watts/ m^2), upwelling IR
case temp. (K), upwelling IR dome temp. (K), global UVB
(milliWatts/ m^2), photosynthetically active radiation (Watts/
m^2), net solar (dw_solar - uw_solar) (Watts/ m^2), net
infrared (dw_ir - uw_ir) (Watts/ m^2), net radiation
(netsolar+netir) (Watts/ m^2), 10-meter air temperature (C),
relative humidity (%), wind speed (ms^1), wind direction
(degrees, clockwise from north), and station pressure (mb).
According to Dobbs [29], Global downwelling solar
measurements best represent the Global Horizontal Irradiance
(GHI) at the SURFRAD sites, which was validated in this
paper through exploratory data analysis. [Figure 3] shows the
daily averages of Clear Sky GHI and global downwelling solar
at SURFRAD site for a year, both the variable follow the same
trend. [Figure 4] shows that both these variables are positively
correlated.
Bird model is used to calculate clear sky GHI [30]. At time
t, clear sky GHI is denoted by ‫ܫܪܩ‬௖௟௘௔௥ ௧
, representing the
theoretical GHI at time t assuming zero cloud coverage. At
time t, the ratio between the instantaneously observed ‫ ܫܪܩ‬௧
௧
and the theoretical maximum ‫ܫܪܩ‬௖௟௘௔௥ is called clear sky
ሺ௧ሻ
index, denoted by ‫ݐܭ‬௜ , this parameter is introduced in [29].
 ሺ௧ሻ quantities, as shown by the regression line plotted on to the
‫ݐܭ‬௜ is used as the dependent variable for training and testing
ሺ௧ሻ scatter plot.
the model. ‫ݐܭ‬௜ is averaged over forecasting horizon, for
hourly predictions. The averaged hourly clear sky index B. Evaluation Metric
ending at time f.h. is denoted by ‫ݐܭ‬௔
ሺ௙Ǥ௛Ǥሻ
and calculated as The algorithm uses Mean Squared Error (MSE) as a
shown in Equation 6. measure to find the difference between the target and the
output that the neural network produces during the training
೑Ǥ೓Ǥ ሺೞሻ process, this is shown in Equation 7. Later in the process, for
σೞస೑Ǥ೓Ǥషలబ ௄௧೔
(6) the purpose of benchmarking, the Root Mean Squared Error is
଺଴ calculated by taking square root of the MSE values.
IV. EXPLORATORY DATA ANALYSIS AND PROPOSED ALGORITHM
ଵ
෡ప ൯ଶ
‫ ܧܵܯ‬ൌ  σ௡௜ୀଵ൫ܻ௜ െ ܻ (7)
A. Exploratory Data Analysis ௡
Where ܻ is a vector of target values and ܻ෠ is a vector of n
predicted values.
C. Algorithm

Figure 2 Cleary Sky and Observed Irradiance

Figure 3 Correlation between Observed and Clear Sky GHI

Figure[2] shows the variation of the observed global
downwelling solar and clear sky global horizontal irradiance
for the year 2010 at Boulder, CO. Figure [3] shows the
correlation between the two for same year and same site.
There is a positive and strong correlation between the two
Figure 4 Flow chart of the proposed method
The flowchart of the proposed Unified Recurrent Neural
Network Architecture based method is shown in Figure 4. The
overall algorithm can be divided into three mail blocks:

21
 1) Preprocessing
The site-specific data is imported and clear sky global
horizontal irradiance values for that site are obtained from the
Bird Model. The two are merged. Dataset is split into train and
testing sets. The clear sky index parameter is created as the
ratio of observed global downwelling solar (Watts /m^2) and
GHI (Watts /m^2). Kt is a dimensionless parameter. The
missing values in the data are replaced by the mean and/or the
neighborhood values. Exploratory data analysis is conducted
to identify and eliminate extreme outliers in order to
normalize the data.
2) RNN training and testing
A Recurrent Neural Network based model architecture is
instantiated by specifying the architectural parameters: input
dimension (number of nodes at the input layer; 22), hidden
dimension (number of nodes in the hidden layer; 15), layer
dimension (number of hidden layers; 1) and output dimension
(number of nodes in the output layer; 4). Sequence length,
which unfolds as time-steps is also defined here along with the
batch size. The model is trained and test by iterating through
the whole dataset based on pre-set number of epochs. We used
a batch size of 100 for 1000 number of epochs for obtaining
the results discussed in this paper.
3) Post-processing
Once the training and testing is over, the stored MSE is
first de-normalized and then it is used to calculate RMSE. If
the RMSE is not satisfactory the hyperparameters (learning
rate and number of epochs) are tuned and the model is trained
again. When a satisfactory (or as expected) RMSE is
achieved, the training process of the algorithm terminates.
V. RESULTS AND DISCUSSION
The algorithm is trained using the data for the year 2010
and 2011 from the SURFRAD observations sites in Boulder,
CO; Desert Rock, NV; Fort Peck, MT; Sioux Falls, SD;
Bondville, IL; Goodwin Creek, MS; and Penn State, PA. The
test year for each respective site was chosen to be 2009 for the
purpose of benchmarking against [29] and other previously
reported results in the literature. Results from the two methods
proposed in this paper are presented in the following two sub-
sections.
Figure 5 Test Mean Squared Error Plot
22
A. Fixed Time Horizon Predictions
The first method uses the proposed RNN architecture and
algorithm to predict for 1-hour, 2-hour, 3-hour and 4-hour
time horizons, independently. In other words, four
independent models are developed for 1-hour, 2-hour, 3-hour
and 4-hour predictions for each of the seven SURFRAD sites.
Figure 5 shows the test loss over 1000 epochs (with a batch
size of 100 and test set of 3300 samples from the test year
2009) for all the seven sites. The RMSE values in Table 1
show that the proposed architecture and algorithm has lower
RMSE values for all four forecasting horizons and all the
seven sites, compared to the best RMSE values reported in
[29] from a suite of other machine learning algorithms
(Random Forests, Support Vector Machine, Gradient Boosting
and vanilla Feed-Forward).
B. Multi-time-horizon prediction
In this method, the architecture predicts for all four
forecasting time horizons (1-hour, 2-hour, 3-hour, and 4-hour)
in parallel; i.e. one model per SURFRAD site is developed
which makes predictions for all the four time horizons. This
method is the multi-time-horizon implementation of the
proposed architecture. None of the methods discussed in the
literature have been shown to be capable of producing multi-
time-horizon predictions. Table 2 enlists the RMSE values
obtained for the test years 2009, 2015, 2016 and 2017. To
quantify the overall performance of the predictive model in
terms of its combined forecasting accuracy for all four
forecasting horizons, the mean of the RMSE values is
calculated. Although, even the best RMSE values reported in
the literature (for example in [29][32]) were for a single time
horizon forecast at a time, the proposed method achieves a
significantly lower RMSE in predictions across all the short-
term (1 hour to 4 hours) forecasting time-horizons as seen in
Table 2.
Capability to predict for multi-time-horizons makes the
proposed method very relevant for industry applications. The
real-time data can be fed to the RNN and due to its lower
forward inference time, predictions can be made for multiple
time horizons. The proposed method is implemented using
PyTorch and the code and additional information can be found
on this site: http://sakshi-mishra.github.io/.
VI. CONCLUSION AND FUTURE WORK
Short-term solar forecasting is of great importance for
optimizing the operational efficiencies of smart grids, as the
uncertainties in the power systems are ever-increasing,
spanning from the generation arena to the demand-side
domain. A number of methods and applications have been
developed for solar forecasting, with some level of predictive
success. The main limitation of the approaches developed so
far is their specificity with a given temporal and/or spatial
resolution. For predictive analysis problems, the field of AI
has become promising with the recent advances in
optimization techniques, parallelism, and GPUs. AI
(especially deep neural networks) thrives on data, and with
decreasing cost of sensor and measurement equipment,
plethora of solar data is getting available. Data availability is
only going to keep increasing in the coming years. The
proposed novel Unified Recurrent Neural Network Although a deeper neural network will have more
Architecture harnesses the power of AI to form a high-fidelity capacity, we experimentally observed that it leads to high
solar forecasting engine. This architecture has the potential to variance in the model and therefore a reduced generalization
be implemented as a complete forecasting system, which power for the particular problem dealt in this paper. The
spans the entire spectrum of spatial and temporal horizons performance of the proposed method can be further improved
with a capability to take real-time data as input to produce in several ways including hyper-parameter tuning and
multi-time-scale (intra-hour, hourly and day-ahead scales) architectural changes like the activation functions used or the
predictions. In addition, the proposed algorithm outperforms type of layers. Extension of the proposed architecture with
traditional Machine Learning methods in terms of quality of LSTM cells and intra-hour forecasting horizons are potential
the forecast and its low forward inference time makes it a future research avenues in this domain.
robust real-time solar forecasting engine.
Table 1
Method 1 (Fixed Time Horizon Predictions) Results
Year 2009 Bondville Boulder Desert Rock Fort Peck Goodwin Penn State Sioux Falls
Creek
F.H. RNN ML RNN ML RNN ML RNN ML RNN ML RNN ML RNN ML
1-hour 16.8 62 17 74 41.7 52 21.2 56 24.8 71 8.64 67 27.2 52
2-hour 20.73 98 20.7 108 57.23 72 29.7 81 25.2 103 10.5 97 32.1 81
3-hour 18.78 116 21.2 123 60.54 83 25.5 94 26.9 125 11.8 114 30.6 96
4-hour 17.98 121 22.9 125 49.71 82 29.4 93 22 120 10.7 117 35.3 103
Mean 18.57 99.25 20.45 107.5 52.29 72.25 26.45 81 24.73 104.8 10.41 98.75 31.3 83
RMSE

Table 2
Method 2 (Multi-time-horizon prediction) Results
Year Forecast Bondville Boulder Desert Fort Peck Goodwin Penn Sioux
Horizon Rock Creek State Falls
1-hour 0.706 0.593 0.469 0.709 0.647 0.679 0.686
2 -hour 1.119 1.055 0.984 1.352 1.173 1.134 1.247
2009 3-hour 4.524 4.010 4.527 6.311 4.586 3.161 5.061
4-hour 52.130 49.453 153.270 69.987 58.984 28.396 73.409
Mean 14.619 13.779 39.812 19.589 16.347 8.342 20.100
RMSE
1-hour 0.717 0.559 0.437 0.689 0.643 0.734 0.680
2-hour 1.166 1.026 0.952 1.302 1.190 1.180 1.252
2015 3-hour 4.719 4.006 4.268 5.845 4.764 3.118 5.216
4-hour 57.839 49.349 85.952 57.687 35.814 25.796 50.823
Mean 16.110 13.735 22.902 16.381 10.603 7.707 14.493
RMSE
1-hour 0.731 0.593 0.456 0.726 0.669 0.738 0.723
2-hour 1.184 1.079 0.979 1.372 1.204 1.187 1.331
2016 3-hour 4.639 4.227 4.685 6.406 4.760 3.174 5.781
4-hour 69.837 89.990 90.553 73.523 45.026 21.994 56.113
Mean 19.098 23.973 24.169 20.507 12.915 6.773 15.987
RMSE
1-hour 0.744 0.593 0.444 0.726 0.649 0.711 0.721
2-hour 1.190 1.059 0.949 1.368 1.179 1.149 1.323
2017 3-hour 4.577 4.010 4.319 6.245 4.593 3.211 5.715
4-hour 81.434 49.453 44.589 114.072 32.431 29.307 59.761
Mean 21.986 13.779 12.575 30.603 9.713 8.594 16.879
RMSE

23
 REFERENCES
1. (2016) The RTO Insider website. [Online]. Available:
https://www.rtoinsider.com/caiso-regulation-requirements-29299/
2. D. Lew, D. Piwko, N. Miller, G. Jordan, K. Clark, and L. Freeman,
“How do high levels of wind and solar impact the grid? The western
wind and solar integration study,” National Renewable Energy
Laboratory, CO, 303, 2010.
3. J. Taylor and P. McSharry, “Short-term load forecasting methods: An
evaluation based on European data,” IEEE Trans. Power Syst., vol. 22,
no. 4, pp. 2213–2219, Nov. 2007.
4. A. Conejo, M. Plazas, R. Espinola, and A. Molina, “Day-ahead
electricity price forecasting using the wavelet transform and ARIMA
models,” IEEE Trans. Power Syst., vol. 20, no. 2, pp. 1035–1042, May
2005.
5. J. Deng and P. Jirutitijaroen, “Short-term load forecasting using time
series analysis: A case study for singapore,” in Proc. IEEE Conf. CIS,
Jun. 2010, pp. 231–236.
6. H. Hippert, C. Pedreira, and R. Souza, “Neural networks for short-term
load forecasting: A review and evaluation,” IEEE Trans. Power Syst.,
vol. 16, no. 1, pp. 44–55, Feb. 2001.
7. A. da Silva and L. Moulin, “Confidence intervals for neural network
based short-term load forecasting,” IEEE Trans. Power Syst., vol. 15,
no. 4, pp. 1191–1196, Nov. 2000.
8. R. Sood, I. Koprinska, and V. Agelidis, “Electricity load forecasting
based on autocorrelation analysis,” in Proc. IJCNN, Jul. 2010, pp. 1–8
9. G. Capizzi, C. Napoli, and F. Bonanno, “Innovative second-generation
wavelets construction with recurrent neural networks for solar radiation
forecasting,” IEEE Trans. Neural Netw. Learn. Syst., vol. 23, no. 11, pp.
1805–1815, Nov. 2012.
10. Q. Yu, H. Tang, K. Tan, and H. Li, “Rapid feedforward computation by
temporal encoding and learning with spiking neurons,” IEEE Trans.
Neural Netw. Learn. Syst., 2013, doi: 10.1109/TNNLS.2013.2245677.
11. Paoli C, Voyant C, Muselli M, Nivet M-L, “Forecasting of preprocessed
daily solar radiation time series using neural networks”, Solar Energy
2010;84(12): 2146e60.
12. Mellit A, Pavan AM, “A 24-h forecast of solar irradiance using artificial
neural network: application for performance prediction of a grid-
connected PV plant at Trieste, Italy”, Solar Energy 2010;84(5):807e21.
13. Chen C, Duan S, Cai T, Liu B, “Online 24-h solar power forecasting
based on weather type classification using artificial neural network”,
Solar Energy 2011;85(11):2856e70.
14. Pedro HTC, Coimbra CFM. “Assessment of forecasting techniques for
solar power production with no exogenous inputs”, Solar Energy
2012;86(7): 2017e28.
15. A. Khosravi, S. Nahavandi, D. Creighton, and D. Srinivasan, “Interval
type-2 fuzzy logic systems for load forecasting: A comparative study,”
IEEE Trans. Power Syst., vol. 27, no. 3, pp. 1274–1282, Aug. 2012.
16. J. R. Castro, O. Castillo, P. Melin, O. Mendoza, and A. Rodríguez-Díaz,
“An interval type-2 fuzzy neural network for chaotic time series
24
prediction with cross-validation and akaike test,” in Soft Computing for
Intelligent Control and Mobile Robotics. New York, NY, USA:
Springer-Verlag, 2011, pp. 269–285.
17. D. Hidalgo, P. Melin, and O. Castillo, “An optimization method for
designing type-2 fuzzy inference systems based on the footprint of
uncertainty using genetic algorithms,” Expert Syst. Appl., vol. 39, no. 4,
pp. 4590–4598, 2012.
18. P. Mathiesen, J.,Kleissl, “Evaluation of numerical weather prediction for
intra-day solar forecasting in the continental United States,” Solar
Energy, vol. 85, pp. 967-977, May 2011.
19. Polo J, Zarzalejo L, RamÕ´rez L, Badescu V, “Solar radiation derived
from satellite images”, Modeling Solar Radiation at the Earth’s Surface,
vol 18. Springer-Verlag Berlin, (ed) 2008.
20. P. Melin, J. Soto, O. Castillo, and J. Soria, “A new approach for time
series prediction using ensembles of ANFIS models,” Expert Syst. Appl.,
vol. 39, no. 3, pp. 3494–3506, 2012.
21. C. Potter and M. Negnevitsky, “Very short-term wind forecasting for
Tasmanian power generation,” IEEE Trans. Power Syst., vol. 21, no. 2,
pp. 965–972, May 2006.
22. Zarzalejo LF, Ramirez L, Polo J., “Artificial intelligence techniques
applied to hourly global irradiance estimation from satellite-derived
cloud index”, Energy; 30(9):1685e97, 2005.
23. M. Noia, C. F. Ratto, and R. Festa, “Solar irradiance estimation from
geostationary satellite data: I. Statistical models”, Solar Energy, vol. 51,
pp. 449-456, Dec. 1993.
24. M. Noia, C. F. Ratto, and R. Festa, “Solar irradiance estimation from
geostationary satellite data: II. Physical models”, Solar Energy, vol. 51,
pp. 457-465, Dec. 1993.
25. R. H. Inman, H. T.C. Pedro, and C. F. M. Coimbra, “Solar Forecasting
methods for renewable energy integration”, Progress in Energy and
Combustion Science, vol. 39, pp. 535-576, Dec. 2013.
26. Y. LeCun, Y. Bengio and G. Hinton, “Deep Learning”, Nature, vol. 521,
pp. 436-444, May 2015.
27. A. Graves, “Supervised sequence labelling with recurrent neural
networks”, Studies in Computational Intelligence, ser. vol. 385, Berlin
Heidelberg: Springer-Verlag, 2012.
28. P. J. Werbos, “Backpropagation through time: what it does and how to
do it.”, Proceedings of IEEE, vol. 78, pp. 1550-1560, 1990.
29. A. Dobbs, “Short-term solar forecasting performance of popular
machine learning algorithms”, National Renewable Energy Laboratory,
Golden, CO, Aug. 2017.
30. R.E. Bird and R.L. Hulstrom, “Simplified clear sky model for direct and
diffuse insolation on horizontal surfaces”, Solar Energy Research Inst.,
Golden, CO. 2013.
31. NOAA Surface Radiation Network (online). Available:
https://www.esrl.noaa.gov/gmd/grad/surfrad/
32. R. Perez, S. Kivalov, J. Schlemmer, K. Hemkar Jr., D. Renne, and T.E.
Hoff, “Validation of short and medium term operational solar radiation
forecasts in the US”, Solar Energy. 84(12), 2161-2172, 2010